From: Howard C. Berkowitz (hcb@xxxxxxxxxxxx)
Date: Fri Feb 15 2002 - 10:58:28 GMT-3
At 2:23 PM +0900 2/15/02, Jaeheon Yoo wrote:
>Hey, Howard.
>I always admire your posts. :)
>But as you admitted at the bottom of your own post, I'm afraid
>you've gone too far here.
You make a very good post. If the immediate goal is to become a CCIE,
the focus needs to be on what is externally visible and controllable,
rather than what goes on inside the code or chips. I hope Cisco
itself finally figured that out and stopped asking to have the major
ASICs on a switch identified!
I was, however, answering some of the original poster's questions
"prove this," when the appropriate answer, at times, is "big magic.
Only witch doctor need to know."
>
>I think that's the difference between the requirements for
>administrators and those for implementors in general.
>When I read some RFC stuffs, I'm always having this kind of confusion.
>Do we ever really need such in-depth knowledge that may be useful
>only to real protocol implementors?
No, but there is useful stuff in RFCs. The trick is knowing how to read them.
In general, if there is an "Applicability Statement" RFC for a
protocol, read the whole thing. That series is meant to explain why
the protocol is there, not how it works.
Next, take an actual protocol specification like RFC2328 for OSPF.
The first few chapters are introductory and worth reading. You'll
then see the level of detail increase tremendously.
When it starts doing that, and you're reading the document for
general information, skip to the end and see if there are any
appendices. The appendices, which are not technically part of the
protocol specification, often deal with implementation details like
timer settings.
Remember these things are often written by teams. I have a BGP
convergence document about to be submitted to the RFC process, and I
can't say I have every definition memorized...I remember the
arguments about some of them, and certainly can glance at a section
and start intelligently discussing it. And this is a document for
which I am the lead author and really have gone through it
word-for-word...but in quite a number of drafts. I might be thinking
of something we never wrote down, or deleted in some draft, etc.
When you have a specific question, yes, by all means search the
document for specific questions.
Also, know the RFCs that are worth reading in full. The RIP
specification is very well written and will teach you a lot. RFC 1812
gives a lot of big picture information.
>Yes, it will help us when we know every detail of ins and outs of
>protocols, I know that.
>But we're already behind the schedule to cover ever expanding
>variety of networking subjects.
>
>Anyway, thanks for your insights.
>
>Jaeheon
>
>----- Original Message -----
>From: "Howard C. Berkowitz" <hcb@gettcomm.com>
>To: <ccielab@groupstudy.com>
>Sent: Friday, February 15, 2002 12:49 PM
>Subject: RE: How route redistribution EXACTLY works (followup)
>
>
>>
>> I hate to follow up my own posts, but let me offer some suggested
>> general reading.
>>
>> Look at RFC1812, which will give some broad specifications.
>>
>> John Moy's OSPF books will give you an example of how a protocol
>> implementation is designed -- the second book has the code of an
>> actual implementation.
>>
>> You can also download the free Zebra code and examine it. Zebra's
>> command language is closer to Cisco's than is GateD. Old versions of
>> GateD are downloadable, but the more recent versions are commercial.
>> Indeed, there's a commercial version of Zebra called IPinfusion.
>>
>> At some point, you're going to need to understand a fair bit about
>> data structures and searching them. Donald Knuth's _The Art of
>> Computer Programming_, Volume 3, has a good deal about it. You also
>> may want to do web searches for radix and Patricia trees (also
>> spelled tries). The IOS internals book has some material on this sort
>> of table.
>>
>> Agreeing with some other posters, this is interesting material, but
> > frankly I'd call it at a level beyond CCIE.
This archive was generated by hypermail 2.1.4 : Thu Jun 20 2002 - 13:46:23 GMT-3